Unsealed documents in The New York Times’ copyright lawsuit against OpenAI and Microsoft reveal startling internal admissions that directly contradict the tech giants’ public legal defenses. A top Microsoft executive, Brent Hecht, privately described the mass scraping of publisher content as “an astonishing theft of unprecedented proportions” and “the largest theft of labor in human history.” These filings, submitted three years into the litigation, expose a stark disconnect between the companies’ fair use arguments and their private assessments of the economic damage inflicted on news organizations.
The evidence suggests AI models are not just transforming content but replacing it. Internal data from Microsoft shows its Copilot “answer engine” drove click-through rates for The New York Times’ domain down by as much as 93% compared to traditional Bing searches. Hecht labeled this decline a “doom loop,” warning that the technology threatens the very “content supply chain” it relies on. OpenAI leadership echoed these concerns, with Nick Turley noting that chatbots are “largely substitutive” and pose an “existential threat” to publishers. Even CEO Satya Nadella testified that paywalled content should be licensed, stating he would have required retraining if he had known OpenAI scraped behind paywalls.
Scale matters here. The filings disclose that OpenAI’s mid-training datasets alone contain over 91,692 copies of works from the Times and other plaintiffs, while a Common Crawl-derived dataset included more than 2 million documents from nytimes.com. The companies allegedly collaborated closely through initiatives like Project Mango and Project Taxi, sharing training data to build commercial products. This level of coordination challenges the notion of independent development and raises questions about who bears liability for the resulting copyright infringement.
While judges have historically leaned toward fair use defenses for AI training, these specific admissions regarding market substitution and deliberate bypassing of paywalls could shift the legal landscape. The Trump administration recently filed a brief supporting OpenAI, but internal memos describing the practice as theft provide powerful counter-narratives. If courts accept that AI outputs directly substitute for original reporting, the fair use shield may crack. The outcome will likely determine whether newsrooms can survive the AI era or become mere data mines for Silicon Valley.
